The fact that works are produced not by a human, not for private use, but by a for-profit tool commercially at scale is important.

> if OpenAI was downloading pirated copies of the Ghibli corpus then that act itself was copyright infringement, independently of the AI training and output.

On the contrary, having legally bought the works is irrelevant. You can buy a Disney movie and be liable if you screen it at a local mall you own.